A rapidly evolved shift in life‐history timing during ecological speciation is driven by the transition between developmental phases

Abstract

For insect species in temperate environments, seasonal timing is often governed by the regulation of diapause, a complex developmental programme that allows insects to weather unfavourable conditions and synchronize their life cycles with available resources. Diapause development consists of a series of distinct phases including initiation, maintenance, termination and post‐diapause development. The evolution of insect seasonal timing depends in part on how these phases of diapause development and post‐diapause development interact to affect variation in phenology. Here, we dissect the physiological basis of a recently evolved phenological shift in Rhagoletis pomonella (Diptera: Tephritidae), a model system for ecological divergence. A recently derived population of R. pomonella shifted from specializing on native hawthorn fruit to earlier fruiting introduced apples, resulting in a 3–4 week shift in adult emergence timing. We tracked metabolic rates of individual flies across post‐winter development to test which phases of development may act either independently or in combination to contribute to this recently evolved divergence in timing. Apple and hawthorn flies differed in a number of facets of their post‐winter developmental trajectories. However, divergent adaptation in adult emergence phenology in these flies was due almost entirely to the end of the pupal diapause maintenance phase, with post‐diapause development having a very small effect. The relatively simple underpinnings of variation in adult emergence phenology suggest that further adaptation to seasonal change in these flies for this trait might be largely due to the timing of diapause termination unhindered by strong covariance among different components of post‐diapause development.

Abstract

The apple maggot fly, Rhagoletis pomonella, has evolved divergent life‐history timing in response to colonizing a novel, introduced host plant in the last ~160 years, resulting in (A) divergent post‐winter metabolic trajectories between the ancestral hawthorn (red) and derived apple (yellow) host races. (B) Despite the multifaceted differences in diapause development between these populations, a single component of diapause explains variation in overall phenology of these flies, suggesting that variation in the transition between developmental modules may allow for the rapid evolution of insect life‐history timing in the face of changing seasonal conditions.

Open in new tabDownload slide

INTRODUCTION

Phenological adaptation can play an important role in both the origin and maintenance of biodiversity through its potential role as potent driver of reproductive isolation during ecological speciation (Taylor & Friesen, 2017) and as a critical axis of evolutionary responses to climate change (Miller‐Rushing, Høye, Inouye, & Post, 2010), respectively. Understanding how evolution shapes the regulation of seasonal timing has important implications for the tempo of ecological diversification, as well as the potential of biotic communities to persist in the face of global climate change. Divergent adaptation to novel habitats or niches can generate reproductive isolation between populations through the process of ecological speciation (Rundle & Nosil, 2005; Schluter, 2001, 2009). Temporal isolation, or allochrony, may be a common component of ecological barriers to gene flow for two reasons. First, temporally distinct resource opportunities may pose distinct trade‐offs, favouring phenological specialization (Berlocher & Feder, 2002). Second, for organisms with seasonal reproductive biology, phenological divergence may serve as a “magic trait” for speciation by pleiotropically coupling ecological divergence to assortative mating (Servedio, Doorn, Kopp, Frame, & Nosil, 2011). Indeed, adaptation to differences in seasonal timing has been shown to contribute to reproductive isolation in a wide variety of organisms including plants (Savolainen et al., 2006; Lowry, Rockwood, & Willis, 2008), insects (Abbot & Withgott, 2004; Eubanks, Blair, & Abrahamson, 2003; Forbes, Powell, Stelinski, Smith, & Feder, 2009; Hood et al., 2015; Horner, Craig, & Itami, 1999; Ording, Mercader, Aardema, & Scriber, 2009; Tauber, Tauber, & Nechols, 1977; Wadsworth, Woods, Hahn, & Dopman, 2013; Wood & Guttman, 1982) and vertebrates (Friesen et al., 2007).

Shifts in phenology have also received increasing attention as anthropogenic climate change causes alterations in the phenology of plant and animal populations (Parmesan, 2006; Thackeray et al., 2016; Van Dyck, Bonte, Puls, Gotthard, & Maes, 2015; Yang & Rudolf, 2010). Many recent studies provide evidence for fitness consequences of disrupting seasonal phenologies, from high bird nestling mortality due to mismatch with seasonal food supplies (Sanz, Potti, Moreno, Merino, & Frias, 2003) to declines in recruitment in plant populations resulting from lack of synchrony between flowering and the activity of pollinators (Gilman, Fabina, Abbott, & Rafferty, 2011; Kudo & Ida, 2013; Rafferty & Ives, 2011). Cases of recent shifts in phenology during incipient speciation also provide useful windows into how the regulation of seasonal timing evolves more generally, offering insight into the factors that may promote or constrain evolutionary responses to anthropogenic change (Scriber, Elliot, Maher, McGuire, & Niblack, 2014; Taylor & Friesen, 2017).

Phenological adaptation is crucial for species with specialist life‐history strategies. For example, univoltine phytophagous insects typically have one chance to breed per year, with most of these species specializing on a single host plant (Strong, Lawton, & Southwood, 1984). Therefore, suboptimal timing can have severe fitness costs (Miller‐Rushing et al., 2010). Many univoltine insects use seasonal dormancy, or diapause, to synchronize their active times for growth and reproduction with seasonal resources. Thus, the decision of when to enter dormancy relative to the timing of unfavourable seasonal conditions, such as the onset of winter, the dry season or lack of host plant availability, as well as when to exit dormancy to synchronize with favourable conditions, is critical aspects of life‐history timing (Tauber & Tauber, 1981). Given that changes to winter conditions are expected to be at the forefront of the environmental impacts of climate change (Williams, Henry, & Sinclair, 2015), the adaptive potential of dormancy‐mediated phenology may be particular important for understanding the consequences for insects in temperate environments.

Insect diapause is not just a simple cessation of development resulting directly from physiologically stressful conditions (i.e., cold‐induced quiescence) (Denlinger, 2002; Hahn & Denlinger, 2011). Rather, diapause is a dynamic, environmentally sensitive developmental programme, consisting of multiple developmental phases: preparation, maintenance and the end of diapause followed by the initiation of post‐diapause development (Koštál, 2006). Each of these developmental phases is characterized by distinct molecular (Koštál, Štětina, Poupardin, Korbelová, & Bruce, 2017; Ragland et al., 2017), morphological (Bryon et al., 2017) and physiological (Ragland, Fuller, Feder, & Hahn, 2009) hallmarks. Selection to produce alternative or divergent phenological strategies may act on the relative duration of, or developmental rates during, these phases (Andrewartha, 1952; Koštál, 2006; Koštál et al., 2017). Whether or not variation in each phase of the diapause and post‐diapause programme represents an independent functional module underlain by distinct genetic architecture or is highly interrelated and covaried has important consequences for the evolvability of seasonal timing (Wagner, Pavlicev, & Cheverud, 2007; Wagner & Zhang, 2011). If independent, then it is possible that each module can be differentially shaped by natural selection to varying environmental conditions at specific points during development. However, if the timing of different phases of diapause development is highly interrelated, then it may be that selection acts on the phase having greatest significance for matching an insect's life cycle with resource availability, with possible pleiotropic consequences on other phases. Although the concept of distinct phases within the diapause developmental trajectory has long been recognized (Andrewartha, 1952; Koštál, 2006; Tauber, Tauber, & Masaki, 1986), how different phases of diapause development interact and how selection acts on variation within these phases to affect life‐history timing are not well understood and have received very little attention despite its potential importance for seasonal adaptation.

To address the question of the flexibility of the diapause programme to compartmentally adapt to environmental challenges requires first characterizing the different phases of diapause development in an insect and then determining their interactions (correlational structure) to assess whether they are assembled into independent modules or not. Under the simplest scenario, seasonal divergence in emergence timing is due principally to the degree of metabolic suppression occurring during the diapause maintenance phase (Feder & Filchak, 1999; Wipking, Viebahn, & Neumann, 1995), with no subsequent significant independent effect of any other developmental phase. Thus, there will be a general 1:1 correspondence between individuals having greater metabolic suppression and later adult eclosion. A competing hypothesis is that the key factor dictating phenology is when diapause maintenance is terminated, regardless of metabolic depth, (Wadsworth et al., 2013). Alternatively, variation in adult emergence phenology may be driven solely by differences in the rate of post‐diapause morphogenesis (Posledovich, Toftegaard, Wiklund, Ehrlén, & Gotthard, 2015; Posledovich et al., 2014). In each of these models, other components of diapause preparation, maintenance and post‐diapause development would be weakly correlated with eclosion time. Conversely, correlated shifts of different components within and between diapause phases can also lead to divergent phenology, especially if driven by directional selection. For example, earlier adult emergence timing may be driven by combinatorial effects of higher metabolic rate during the diapause maintenance phase, earlier termination of the diapause maintenance phase and more rapid post‐diapause development.

Here, we examine the role that these different components of diapause development play in the case of rapidly evolved phenological divergence in the apple maggot fly, Rhagoletis pomonella (Walsh), a text‐book system for ecological speciation (Berlocher & Feder, 2002; Dres & Mallet, 2002). Divergent seasonal adaptation of the flies to synchronize adult emergence with their specific host fruits is well understood (Feder & Filchak, 1999), making R. pomonella an excellent system for studying how different components of diapause development interact in the regulation of seasonal timing (Dambroski & Feder, 2007; Ragland et al., 2017). Partially reproductively isolated populations that show consistent patterns of host‐associated genetic differentiation, hereafter host races, are undergoing divergent selection to synchronize themselves with the fruiting times of their respective host plants. The derived host race shifted from infesting fruits of a native plant, the downy hawthorn (Crataegus mollis) to infest domesticated apples (Malus pumila) ~160 years ago (Bush, 1969; Walsh, 1861). The apple varieties favoured by R. pomonella ripen ~1 month earlier in the season compared to hawthorns in the Midwestern United States (Feder, Hunt, & Bush, 1993; Feder et al., 1994). As short‐lived, univoltine specialists, apple and hawthorn flies spend most of the year in pupal diapause to synchronize their life cycles with their respective hosts. One well‐characterized axis of divergence in this system involves the successful initiation and maintenance of diapause under extended periods of warm prewinter conditions (Dambroski & Feder, 2007; Egan et al., 2015; Filchak, Feder, Roethele, Stolz, & Mallet, 1999). However, the difference in adult and larval phenology between the races is determined by how flies transition from the shared state of successful pupal diapause during winter to completing adult development prior to emergence at different points the following summer and fall.

We investigate how variation in components of diapause maintenance and post‐diapause development is associated with shifts in seasonal life‐history timing in the derived (apple) compared to the ancestral (hawthorn) host race of R. pomonella. Specifically, we test whether the earlier phenology of the derived apple race is driven by alteration to individual components or multiple components of their post‐winter diapause developmental trajectories. We use respirometric phenotyping to compare the post‐winter metabolic rate trajectories of apple and hawthorn flies from sympatric populations under controlled laboratory common‐garden conditions to identify the key phases of diapause development. Both apple and hawthorn host races are univoltine specialists, whose adult eclosion phenology must coincide with host fruit availability long after the cessation of cold temperatures, in late summer and early autumn, respectively (Bush, 1969). Thus, adaptive seasonal timing requires both host races to maintain diapause development through the transition for cold to physiologically permissive temperatures in spring. Low metabolic rate during warm, post‐winter conditions is a reliable marker for the diapause maintenance phase that identifies individuals in diapause; metabolic rate then increases in a stereotypical trajectory during post‐diapause development following the end of diapause (Ragland et al., 2009). We follow cohorts of flies from both host races in the laboratory from the cessation of simulated winter through to adult emergence, producing a time series of metabolic rates for each individual. We fit these time series to mathematical models with parameters describing different aspects of diapause development corresponding with post‐winter metabolic trajectories that include parameters modelling: (a) baseline metabolic rate during diapause maintenance, (b) timing of the transition from maintenance to post‐diapause development, (c) metabolic rate at a plateau coinciding with early adult morphogenesis, (d) duration of the plateau and (e) an exponential increase in metabolic rate corresponding with late adult morphogenesis culminating in adult emergence (Figure 1). Comparisons across these individual models of metabolic trajectories allow us to test how patterns of post‐winter diapause development differ between the two host races and importantly test which component or combination of components is significantly associated with variation in adult emergence phenology. Thus, our approach provides a test of the hypothesis that variation in the duration of one or multiple components of the diapause maintenance phase or post‐diapause development accounts for earlier emergence timing in the derived apple host race. Because immediate post‐winter metabolic rates were found to be more dynamic than expected, we also conduct a follow‐up experiment comparing a more fine‐scaled time series of overwintering and immediate post‐winter metabolic dynamics between the host races and testing how variation in these traits influences adult emergence timing.

FIGURE 1

Examples of plotted metabolic trajectories following the (a) monophasic, (b) biphasic and (c) triphasic models described in the text. To demonstrate how different parameters affect aspects and the shape of metabolic trajectories and reflect different modules of diapause development, the values of the fitted parameters in each model are indicated in the plots with arrows pointing to their relevant position

Open in new tabDownload slide

MATERIALS AND METHODS

Insect collection and rearing

In the summer and fall of 2013 and 2014, apple and hawthorn race R. pomonella flies were collected from infested fruit from wild populations in Urbana, IL, during peak infestation for both host plants. The apple and hawthorn sites are located ~1,300 m apart, well within the flight range of these flies (Roitberg, Cairl, & Prokopy, 1984). Infested fruit were collected from the ground and trees and transported to the laboratory where they were kept in an environmentally controlled room under standard Rhagoletis rearing conditions (24°C and 14:10 light: dark (L:D): Filchak, Roethele, & Feder, 2000; Dambroski & Feder, 2007). Fruit were placed in wire mesh baskets in plastic trays where wandering third‐instar larvae emerged naturally from the fruit and dropped into the trays. Trays were monitored daily for larvae, which were placed in Petri dishes containing moist vermiculite to pupariate. Petri dishes were kept at 24°C and 14:10 L:D for 10 days. At the end of the 10 days, pupae were transferred to clean Petri dishes and placed in chambers with 85% humidity maintained by a saturated KCl solution (Winston & Bates, 1960). Ninety‐six pupae were selected using a random number generator from the intact pupae across three temporal blocks covering a 12‐day window of peak larval emergence from each host race. Pupae were placed into three measurement groups of 36 individuals each at days 1, 6 and 12 after daily larval emergence exceeded 100 individuals. Humidity chambers were then transferred to simulated winter conditions at 3.5°C ± 1.0°C, where they remained for 24 weeks. The overwintering period was chosen because it allows for adequate chill length for potentiating diapause termination and largely precludes the survival of pupae that either failed to initiated diapause or prematurely terminated diapause prewinter (Feder, Powell, Filchak, & Leung, 2010; Feder et al., 1997). Moreover, this overwintering length is consistent with other recent studies on the transcriptomic and genetic basis of diapause termination in this system (Doellman, Egan, et al., 2018; Meyers et al., 2016; Ragland, Egan, Feder, Berlocher, & Hahn, 2011; Ragland et al., 2011, 2017). After removal from the cold, pupae were kept at 25°C, 14:10 L:D and 85% humidity for the duration of the experiment.

Measuring post‐winter metabolic trajectories

Metabolic rates were measured using stop‐flow respirometry (Lighton, 2008), closely following the methods of Ragland et al. (2009). Upon removal from the cold, 96 pupae from each host race were weighed and placed into individual 5‐ml Norm‐Ject™ syringes (Air‐Tite Products) fitted with three‐way luer valves (Cole‐Parmer). Rhagoletis pupae are small enough (~5–11 mg) to reside within the top arm of the luer valve so that the syringe piston can be fully depressed. Syringes were flushed with CO₂ free air, using an aquarium pump pushing air through a Drierite™ (W.A. Hammond Co.)–Ascarite II™ (Thomas Scientific)–Drierite™ scrubber column followed by an acidified (pH < 4.0) water bubbler to rehumidify the air. Syringes were sealed with 1 ml scrubbed air, and the time of the syringe purge was recorded using Expedata logging software (Sable Systems International). Four empty control syringes were also purged and sealed. Syringes were then returned to 25°C, 14:10 L:D incubators. After approximately 24 hr (exact time recorded for each individual), the full volume of the 1‐mL chamber was injected into a flow‐through system with a Li‐Cor 7000 infrared CO₂ analyser (Lincoln, NE, USA) interfaced with Expedata. The flow rate of the respirometry system was maintained at 150 ml/min using an MFC‐2 mass flow control unit (Sable Systems International) with a Sierra Instruments mass flow valve, and the circulating air in the system was scrubbed of CO₂ and water using a Drierite™ column followed by a Drierite™–Ascarite II™–Drierite™ column. Water vapour from the injected sample was removed by an additional magnesium perchlorate column before entering the sample cell of the gas analyser. Because both the purge and injection times were recorded in Expedata to the second, the exact time that each pupa spent sealed in the chamber was known and used in the subsequent analysis of respiration rates. After injection of the sample air, syringes were returned to the humidity chambers with the luer valves open for 24 hr before the measurement cycle began again, resulting in a metabolic rate measurement every 48 hr until adult emergence. Calibration of the gas analyser was maintained during the experiment using pure nitrogen and a certified mixture of 465.8 ppm CO₂ in nitrogen (Airgas).

The respiration rate for each individual at each time point was calculated using the manual bolus integration method of Lighton (2008) with

where R is the rate of respiration in μl CO₂/hr, C is the instantaneous concentration of CO₂, F is the flow rate in l/s, e is the elapsed time interval in seconds between syringe purge and injection, and t_i‐t_f is the time interval in seconds over which the injection bolus is detected by the gas analyser. We conducted subsequent analyses on wet‐mass equivalent metabolic rates in μl CO₂/hr/mg. Note that the expected variation in developmental state among individuals at a given time point precluded us from establishing a more complex allometric relationship between body mass and metabolic rate in this system.

Analysis of metabolic trajectories

The previous study by Ragland et al. (2009) described a biphasic shape to R. pomonella post‐winter metabolic trajectories, with a depressed baseline respiration rate during diapause punctuated by a logistic increase in CO₂ production following the end of diapause that stabilizes at a plateau level before entering an exponential increase phase during later pharate adult development (Figure 1b). Initial inspection of our data revealed that this basic biphasic pattern was present, but the majority of individuals in our study also revealed a characteristic increase in metabolic rate after transfer from cold to warm conditions, followed by an attenuation back down to a stable baseline (Figure 1c). This initial increase in metabolic rate followed by a return to depression 2–3 days after, indicating pupae were still in the diapause maintenance phase, was also confirmed by an additional follow‐up experiment described below. To account for this additional feature, we fit and compared three nested nonlinear models to each individual fly's time series of respiration rates: (a) a monophasic model consisting of an exponential increase from a single post‐winter baseline metabolic rate (Figure 1a):

(b) a biphasic model similar to the one described by Ragland et al. (2009), but redesigned so that the parameter values are more directly reflective of landmark time points and metabolic rates (Figure 1b):

and (c) a triphasic model that also includes an initial post‐winter elevated respiration rate followed by an immediate exponential decrease down to a baseline diapause rate (Figure 1c):

where R is the mass equivalent respiration rate at time t, b is the baseline post‐winter metabolic rate, p is the plateau metabolic rate between diapause termination and pharate adult development, X represents the timing of the exponential increase in respiration during pharate adult development (equal to the time at which the metabolic rate is 2× higher than the previous stable rate b or p), m is the timing of the logistic increase during diapause termination, i is the initial elevated post‐winter metabolic rate, and c is a scaling factor for the slope of the exponential phase. Initial AIC analysis during model development did not support the inclusion of additional scaling parameters for the logistic increase or initial decrease phase for any individuals.

All three models were fit for each individual in R 3.0.1 (R Development Core), using the nls function. Starting values for each parameter were obtained from visual inspection of the metabolic trajectories, each parameter except for c models either a metabolic rate or time, and thus a position on the y‐ or x‐axis of the metabolic trajectory plot, respectively (Figure 1). Parameters were constrained to being biologically reasonable (e.g., all time and metabolic rate parameters > 0). Models were compared using AICc in the R package MuMIN (Barton, 2015). Goodness of fit of the models selected by AICc was assessed by examining fitted curves and the distribution of residuals.

Testing whether variation in adult emergence timing is associated with variation in the same developmental trajectory model parameters across both host races

If particular phases of diapause or post‐diapause development account for differences in adult emergence phenology, then interindividual variation in parameters from the above models representing these phases should associate with adult emergence timing (a) within and (b) between fly host races. Due to strong covariance among diapause trajectory parameters, as well as collinearity between some parameter estimates and their errors (as expected in this case where the degrees of freedom in each trajectory model vary exactly with phenology because flies that take longer to emerge are measured more times), we relied primarily on a partial correlation approach rather than multiple linear regression for statistical testing. This approach measures the strength of association between variables after controlling for the effect of shared correlations with other variables. We conducted partial correlation analyses assessing the relationships of the parameters b, m, p, X‐m, c to days to adult emergence using the R package ppcor (Kim, 2015). Analysis of variation was conducted both across the host races considering host as a binary variable, and within apple and hawthorn flies considered separately.

Analysis of overwintering and immediate post‐winter metabolic rates

The unexpected presence of elevated metabolic rates in the first few post‐winter measurements 2–3 days after removal from cold combined with early diapause termination of many apple flies introduced more variability into the initial post‐winter baseline measurements than expected in the original experiment. To better control for the rapid change in metabolic rates during this immediate post‐winter phase and to further examine the dynamics of the immediate post‐winter metabolic increase and subsequent decline, we performed a second metabolic phenotyping experiment the following year. Here, we measured metabolic rates of 40 apple race and 40 hawthorn race pupae from the same populations at Urbana, Illinois (but collected during the 2014 field season). Larvae and pupae were handled as above except that we used a 28‐week simulated winter (due to availability of pupae relative to other ongoing experiments). At the end of this 28‐week cold period, we placed the pupae in syringes, purged the syringes with CO₂ free air as described above and sealed them under continuous simulated winter conditions (3.5°C) for 24 hr. After 24 hr, syringes were placed in an insulated cooler with ice packs and brought to our room temperature respirometry system. Sealed syringes were kept in the cooler until immediately before injection, resulting in a measurement of CO₂ produced by each individual under uninterrupted winter temperatures. The post‐winter metabolic rate of the flies was then measured for 12 days at 24°C as described above with the exception that measurements were made every 24 hr. Syringes were repurged immediately after injection. Metabolic rates were calculated as described above. Pupae were then kept in humidity chambers in 0.2‐ml perforated tubes and monitored daily for adult emergence.

Statistical analyses of immediate post‐winter metabolic rates

Our analyses of this second experiment had two distinct goals: (a) to determine whether the two host races differ in their metabolic profiles over this dynamic period of immediate post‐winter metabolic change and (b) to determine the extent to which metabolic rates during the overwintering and immediate post‐winter phase may influence the timing of adult emergence. To test whether and at which time points the host races differ in their immediate post‐winter diapause metabolism, we conducted a repeated‐measures ANOVA of daily metabolic rates for each fly as a function of host race to determine metabolic time points that differed significantly between apple and hawthorn flies. Based on those results, we then asked how variation in two key post‐winter time points that differentiated the host races, specifically 1 and 10 days out of simulated overwintering, and variation in the cold overwintering metabolic rates covaried with each other and with adult emergence phenology using partial correlation analyses, as above.

RESULTS

Metabolic rate trajectories suggest stereotypical developmental trajectories

Of the initial 96 pupae for each host race, 31 apple flies and 35 hawthorn flies survived to adult emergence and were used in subsequent analyses (others either died inside their puparia, were killed during daily handling or were parasitized by braconid wasps). Apple host race flies emerged as adults significantly earlier than hawthorn host race flies with mean and median eclosion times for the apple race flies at 51 and 52 days (SE = 2.33, IQR = 38.25–56.75), respectively, compared to 81 and 79 d (SE = 2.98, IQR = 73–87.5) for the hawthorn race (Figure 2a, KS test, D = 0.793, p < .00001). The majority of pupae followed either a biphasic or triphasic metabolic rate trajectory model based on AICc model selection (Figure 2b). However, there was a single fly showing the monophasic trajectory that emerged only 19 days after simulated winter. AICc results for each fly are presented in Table S1, and the best‐fit models for each fly are presented in a file in the online supplementary material and Figure S1. Apple flies had more metabolic rate trajectories following the biphasic model fit, whereas hawthorn flies had a roughly equal ratio of bi‐ and triphasic model fits (G test, G₂ = 16.962, p < .001), Table S1). The lower frequency of triphasic trajectories in apple flies was likely driven by their shorter time to adult emergence. With fewer data points during the baseline metabolic rate phase of the trajectories, apple flies tended to have less power to fit the initial decrease curve, with the logistic increase of diapause termination often beginning before the decrease phase stabilized (Figure S1).

FIGURE 2

Plots showing (a) that apple flies (yellow) emerge as adults earlier than hawthorn flies (red) and (b) the best‐fit nonlinear models of the metabolic trajectories for each apple fly (yellow) and hawthorn fly (red) that survived to adult emergence in experiment 1

Open in new tabDownload slide

The pattern of initial decrease of metabolic rates from an immediate post‐winter peak was present in both hawthorn and apple flies (Figure S1 and see results of second experiment below). Curves from the fitted models (Figure 2b) generally fit the shape of the data closely (Figure S1). The mean residual standard errors S were low for the best‐fit models, with a mean of 0.0041 μl CO₂/hr across all flies and a range of 0.0012–0.0089 μl CO₂/hr. We note that R² is generally not an appropriate metric of goodness of fit for nonlinear regressions (Spiess & Neumeyer, 2010). To use parameter estimates from the best‐fit models for each fly in subsequent analyses, we tested whether the subset of nested shared parameter estimates differed between the hierarchical biphasic and triphasic models fit for the same individuals. The shared parameters: baseline metabolic rate during the diapause maintenance module (b), timing of the end of the diapause maintenance module (m), post‐diapause plateau metabolism (p), post‐diapause plateau duration (X‐m) and exponential scaling (c) did not differ between the bi‐ and triphasic models in paired t tests (Table S2).

Apple and hawthorn flies differ in key post‐winter metabolic trajectory parameters

Although the host races share the same basic metabolic trajectory templates, apple and hawthorn flies differed in parameters related to both phase‐specific metabolic rates and, most notably, the timing of the transition out of the diapause maintenance phase. Estimates of three parameters: baseline metabolic rate b, the timing of the transition out of the diapause maintenance phase m and the post‐diapause plateau metabolic rate p, differed between host races (Figure 3a–c; two‐tailed t tests, t_36.75 = 4.0787, p = .0026; t_60.98 = −8.61, p = 3.64 × 10⁻¹²; t_51.2 = 3.8012 p = .000385, respectively). Neither the relative timing of exponential metabolic increase during adult development X‐m nor the exponential scaling parameter c were significantly different between host races (Figure 3d,e; t tests, t_62.2 = −0.39081, p = .6973; t_53.75 = 1.444, p = .1542, respectively). Compared to hawthorn flies, apple flies had higher baseline metabolic rates, earlier diapause termination times and higher post‐diapause metabolic plateaus (Figure 3a–c).

FIGURE 3

Box plots comparing fitted parameter values that show: (a) apple flies have a higher baseline metabolic rate (b) than hawthorn flies, (b) apple flies have earlier timing of diapause termination (m) than hawthorn flies, (c) apple flies have higher post‐diapause plateau metabolic rates (p) than hawthorn flies, (d) apple and hawthorn flies do not differ in post‐diapause plateau duration (X‐m) nor (e) exponential scaling (c). The box plots are followed by scatter plots of the relationships between adult emergence time on the y‐axis and (f) baseline metabolic rate (b), (g) timing of diapause termination (m), (h) post‐diapause plateau metabolism (p), (i) post‐diapause plateau duration (X‐m), and (j) exponential scaling (c) with individual apple flies indicated in yellow and hawthorn flies in red, showing that adult emergence timing is strongly correlated with the timing of diapause termination (r = .994) with much weaker relationships to the other fitted metabolic trajectory parameters

Open in new tabDownload slide

Associations between model parameters and emergence timing within host races

Our correlation analysis revealed a strong relationship across both host races between m, the timing of the end of the diapause module and adult emergence timing across both host races (r₆₃ = .994, p = 1.07e⁻⁵⁷, Figures 3g and 4). Earlier timing of the end of the diapause maintenance module was tightly correlated with earlier emergence timing (Figure 3g). Weaker but significant partial correlations with adult emergence timing also existed between post‐diapause plateau metabolism (p), metabolic rate duration of the post‐diapause plateau (X‐m) and the slope of the exponential increase phase c (r₆₃ = −.37; p = .0013; r₆₃ = .772, p = 5e⁻¹³; and r₆₃ = −.533, p = .00015, respectively; Figures 3i,j and 4). The baseline metabolic rate b did not show significant direct correlations with adult emergence time (Figures 3f and 4). These patterns were consistent both within and across the two host races (Figures 3f–j and 4; Figure S2). The strength of the correlation between m and adult emergence timing was nearly identical when the partial correlation analyses were run separately for each host race (apple r = .98 and hawthorn r = .99; Figure S2), reinforcing the importance of the timing of the end of the diapause maintenance module as the primary driver of phenological divergence between the apple and hawthorn host races.

FIGURE 4

Partial correlation network showing the major contribution of the timing of diapause termination (m) along with the minor contribution of post‐diapause parameters describing the plateau metabolic rate (p), the duration of the post‐diapause plateau (X‐m) and scaling of the exponential phase (c) with adult emergence timing as well as additions significant partial correlations among parameters. Edges depict all significant (p < .05) partial correlations. Edge weights are proportional to the magnitude of Pearson's r, for both positive correlations (solid lines) and negative correlations (dashed lines)

Open in new tabDownload slide

Host races diverge in how they respond to initial favourable conditions

The overall shape of the metabolic trajectories for both host races in the second respirometry experiment was consistent with the dynamics found in our first experiment. The end of simulated winter and transfer to 25°C resulted in a sharp initial metabolic increase that peaked at 24–72 hr after removal from the cold and subsequently declined to a post‐winter baseline metabolic rate over the course of 10 days (Figure 5). Overwintering metabolic rates (measured in the cold) were nearly an order of magnitude lower than post‐winter baseline metabolic rates from the experiment above, mean 0.00115 versus 0.01035 μlCO₂/hr/mg, respectively, and did not differ between the host races (Figure S3). However, the two host races did differ in their immediate response to warming, with apple flies being more metabolically responsive to warming than hawthorn flies (repeated‐measures ANOVA: time × host, F_12,552 = 3.54 p < .001; Table S3; Figure 5). Thus, apple flies were more metabolically responsive to the shift from cool simulated overwintering conditions to warm post‐winter conditions, rising to higher metabolic rates than hawthorn flies after simulated overwintering, particularly on days 1 and 2 post‐winter, with some individuals appearing to terminate diapause before day 10 (Figure 5).

FIGURE 5

Plots of metabolic rate every 24 hr over 12 days for surviving apple flies (yellow) and hawthorn flies (red) from experiment 2 showing the immediate post‐winter metabolic increase over days 1 and 2 exhibited in both host races, with apple fly pupae increasing their metabolic rates substantially higher than hawthorn fly pupae after removal from the cold. After this initial metabolic increase, all hawthorn pupae and most apple pupae enter back into metabolic suppression at baseline post‐winter diapause maintenance levels. Time zero on the x‐axis represents the overwintering metabolic rate measurement over which flies were purged and measured under continuous simulated winter conditions. See Figure S4 for a box plot depicting variation in overwintering metabolic rates in more detail

Open in new tabDownload slide

Partial correlation analysis of overwinter, day 1 and day 10 metabolic rates with adult emergence time in the second experiment showed significant but weaker overall covariance among variables than was found in the first experiment (Figure 6). Overwintering metabolic rate was weakly but significantly positively correlated with metabolic rates at the two other time points day 1 and day 10 after removal from cold, as well as host (r₅₁ = .37, p = .008, r₅₁ = .346, p = .013; r₅₁ = −.315, p = .026, respectively, Figure 6 and Figure S5). The only variable with a direct correlation to adult emergence timing was an inverse relationship with day 1 metabolic rate (r = −.318, p = .025, Figure 6 and Figure S5). Pupae of both host races that were more metabolically responsive during the rapid rise in metabolism after shifting from cool overwintering conditions to warm post‐winter conditions were more likely to emerge earlier as adults. This correlation was much weaker than that observed between adult emergence timing and the timing of the end of the diapause maintenance module (r = −.318 vs. r = .994), indicating that if day 1 metabolic rate influences adult emergence phenology, it likely does so through its effect on the timing of the end of diapause maintenance.

FIGURE 6

Partial correlation network showing the relatively weak negative correlation between day 1 metabolic rate and adult emergence timing as well as additional significant partial correlations among overwintering metabolic rate and post‐winter metabolic variables in experiment 2. The three metabolic rate variables over this period are modestly positively correlated, and pupae with particularly strong metabolic increases at day 1 tended to emerge earlier as adults. Edges depict all significant (p < .05) partial correlations among variables. Line weights are proportional to the magnitude of Pearson's r, for both positive correlations (solid lines) and negative correlations (dashed lines)

Open in new tabDownload slide

DISCUSSION

Diapause is an important component of adaptation to temperate environments for most insect species, and it may often serve as a critical step governing the seasonal phenological timing of insect life histories (Tauber & Tauber, 1981). As a multifaceted, alternative developmental pathway, diapause is not just a static resting state. Diapause is a complex trait, possessing a number of physiologically distinct functional components that are expressed along the course of diapause development (Koštál, 2006). Thus, the ability for insect populations to rapidly evolve new phenologies is likely dependent on the complexity with which these different components of diapause development interact to affect overall seasonal timing. Here, we show that although apple and hawthorn flies differ in a number of timing and metabolic rate parameters of post‐winter development (Figure 3a–e), variation in adult emergence, both between and within the host races, is primarily due to differences in the timing of the transition between diapause maintenance and post‐diapause development (parameter m; Figures 3f–j and 4). In both host races, adult flies emerged on average 31.6 days after diapause termination, with a standard deviation of 2.4 days. Wadsworth et al. (2013) reported similar results for a primary effect of diapause termination on adult emergence time in the univoltine‐Z and bivoltine‐E strains of the European corn borer, another system where divergence in seasonal phenology drives allochronic isolation (Dopman, Robbins, & Seaman, 2010), perhaps highlighting the general importance of diapause termination timing in the evolution of insect phenology.

In addition to diapause termination, adult emergence timing in the host races is related, although to a much lesser degree, to two other post‐winter development parameters: (a) the duration of the post‐diapause plateau (X‐m) when early pharate adult morphogenesis is occurring (Ragland et al., 2009) and (b) the slope of the exponential increase phase of pharate adult development (c) late in the process of adult morphogenesis (Figure 3i,j). In our second experiment, a weak but significant inverse correlation was also observed between adult emergence and the initial day 1 post‐winter metabolic rate, such that pupae that more strongly increase metabolic rates after the transition from cool overwintering temperatures to warm post‐winter temperatures are more likely to emerge as adults earlier (Figure 6). However, the timing of the end of the diapause maintenance phase appears to be the major factor driving variation in overall adult emergence phenology, with the later‐acting post‐diapause development parameters having much subtler effects on any residuals variation in emergence timing (Figures 3h–j and 4).

The overall covariance structure presented in the partial correlation network (Figure 4) indicates that the major effect of the timing of the transition between the diapause maintenance and post‐diapause development phases on phenology is not entirely independent of the other metabolic trajectory parameters. Although the timing of transition between phases appears to be the primary driver of phenology, this variation may not act in a singularly modular fashion, as we would infer if the parameters correlated with adult emergence timing did not otherwise overlap in the partial correlation network. However, most of the correlations among metabolic trajectory parameters were relatively modest (only two were stronger than r = .5, with the strongest being between m and X‐m at r = −.76), implying that their respective responses to selection may not be highly constrained. Thus, the covariance between diapause termination and post‐diapause development is not absolute. Both the diapause maintenance phase and the post‐diapause development phase may be capable of separate responses to selection on adult life‐history timing. This result that has broader implications for phenological adaptation in many contexts from seasonal synchronization with novel hosts to rapid adaptation to new climatic regimes during introductions or range expansion to shifting seasonality as a result of climate change (e.g., Bradshaw & Holzapfel, 2001; Gomi, Nagasaka, Fukuda, & Hagihara, 2007; Scriber & Ording, 2005; Wadsworth et al., 2013).

Our focus here was specifically on how components of post‐winter diapause development drive variation in phenology: How do these flies differ in their transition out of a shared state of overwintering dormancy to produce divergently adaptive adult emergence phenologies? The current study does not directly address the well‐studied variation in prewinter diapause initiation and maintenance driven by gene by environment interactions in this system (Dambroski & Feder, 2007; Egan et al., 2015; Feder et al., 2010). Previous work in R. pomonella has identified classes of prewinter diapause phenotypes expressed under certain conditions, whose frequencies differ between sympatric apple and hawthorn flies, including “shallow diapausing” flies that break dormancy under prolonged warm conditions and “chill dependent” flies that require overwintering to break dormancy (Dambroski & Feder, 2007). Recent genomic studies suggest that variation in the pre‐ and post‐winter aspects of diapause divergence are largely independent in this system (Ragland et al., 2017). The combination of rearing conditions chosen for these experiments—short prewinter periods and relatively long overwintering periods—means that any flies that did not successfully initiate or maintain diapause were unlikely to survive to adult emergence. Even if there were substantial differences in the proportion of shallow diapausing and chill dependent individuals between the populations in these experiments, our results show no evidence of bimodality in the strong relationship between the timing of diapause termination and adult emergence phenology (Figure 3g). Nevertheless, it is possible that some of our results may be specific to the specific overwintering conditions of these experiments, with more influences from prewinter conditions if flies are exposed to shorter simulated winters.

In certain insects, variation in diapause phenotype may have relatively simple genetic architecture. Diapause‐mediated allochronic isolation in the European corn borer is under the control of a segregating locus of major effect on the Z chromosome (Dopman, Pérez, Bogdanowicz, & Harrison, 2005; Glover, Robbins, Eckenrode, & Roelofs, 1992; Wadsworth, Li, & Dopman, 2015). Similarly, variation in diapause induction appears to be driven largely by a large‐effect Z‐linked locus in a butterfly species complex comprised of Papilio glaucus, P. canadensis and P. appalacheinsis (Hagen & Scriber, 1989; Kunte et al., 2011). But latitudinal variation in life‐history timing in Papilio butterflies may also be associated with autosomal regions as well (Ryan et al., 2017). In contrast, geographic variation in photoperiodic sensitivity for diapause induction in the mosquito Wyeomyia smithii appears to be under more complex genetic control, involving multiple loci of moderate effect with strong epistasis (Lair, Bradshaw, & Holzapfel, 1997; Mathias, Jacky, Bradshaw, & Holzapfel, 2007). Genetic variation in life‐history timing in R. pomonella appears to be highly quantitative, showing widespread statistical associations with markers across the genome (Egan et al., 2015; Filchak et al., 2000; Michel et al., 2010; Ragland et al., 2017). Our results suggest that the many genes implied to affect adult emergence phenology primarily do so by their effect on the timing of termination of the diapause maintenance module. Thus, timing of diapause termination is a developmental hub through which quantitative allelic variation flows to modify adult emergence phenology.

The genetic complexity of life‐history timing in R. pomonella might seem at odds with the relative simplicity of diapause termination primarily dictating adult emergence timing. However, there is good reason to suspect that many genes or gene networks may play a role in driving variation in the duration of the diapause maintenance module. The prewinter initiation of diapause is not simply a switch ceasing development, and the termination of diapause is not simply a switch resuming development. Rather the phases are each characterized by a progression of physiological changes that lead up to each transition along the developmental pathway (Denlinger, 2002; Koštál, 2006; Koštál et al., 2017). The dynamic nature of this developmental trajectory allows for many different transitions at which allelic variation might influence diapause timing, and many intermediary developmental avenues exist along which highly polygenic variation could act additively to affect eclosion timing. Precisely how each of these processes and events relates to the detectable metabolic increase during diapause termination is not currently known. Further characterizing variation in these intermediate mechanistic steps to diapause termination will help parse the complex pattern of genomic variation associated with life‐history timing in these flies.

The earlier diapause termination of apple than hawthorn flies may partially be due to differences in their developmental progression during the overwintering portion of the diapause maintenance phase. Apple pupae may become more primed or potentiated to respond to environmental cues sooner after winter than hawthorn pupae. For example, apple flies may be prone to enter or progress further into the stage of diapause development permissive for termination of the maintenance phase earlier in the winter compared to hawthorn flies. Alternatively, the host races could emerge from winter in the same state of diapause development, with the difference in adult emergence time resulting from variation in their respective responses to subsequent environmental cues. Although the two host races did not show a detectable difference in overwintering metabolic rate, the fact that apple flies were more metabolically responsive to the shift from cool overwintering temperatures to warm post‐winter temperatures, and the very rapid nature of diapause termination in many of the apple flies, argues that apple fly pupae are potentiated to respond to diapause termination cues earlier than hawthorn fly pupae. The finding of higher initial metabolic response to warming in apple flies combined with the association between earlier adult emergence and higher initial metabolic rates just after warming in both host races suggests a modification to the simple hypothesis that basal metabolic rates during diapause themselves are a primary driver of diapause duration (Feder & Filchak, 1999; Wipking et al., 1995). Instead, perhaps the dynamic responsiveness of metabolism to thermal shifts indicates greater potentiation to respond to diapause termination cues both within and between the host races and acts upstream of the timing of the end of diapause maintenance.

A recent RNAseq study comparing the transcriptomes of the host races during simulated overwintering and soon after removal from the cold is consistent with the hypothesis that apple flies are more potentiated to terminate diapause after winter than hawthorn flies (Meyers et al., 2016). The head transciptomes (including neuro‐endocrine tissues) of apple and hawthorn flies were found to already be strongly differentiated before pupae were removed from the cold, with biases towards up‐regulation of genes involved in cell proliferation and development in the apple flies (Meyers et al., 2016). Our finding of greater responsiveness of metabolic rate in apple than hawthorn race pupae as they shift from cold, developmentally suppressive temperatures to warm, developmentally permissive temperatures also suggests apple race pupae become more potentiated for diapause termination by the end of simulated winter treatments compared to hawthorn race pupae.

The immediate post‐winter metabolic increase and subsequent decline observed in our second experiment (Figure 5) are a novel finding with implications for how dormancy is maintained in the face of abiotic conditions that are otherwise permissive for nondiapause development. The physiological mechanisms underlying these dynamics remain to be determined. The metabolic increase may be similar to infradian metabolic cycling observed in diapausing pupae of other species (e.g., Sláma & Denlinger, 1992). Alternatively, these dynamics could be related to the release of CO₂ built‐up over winter if warm conditions trigger changes to discontinuous gas exchange (e.g., Wobschall & Hetz, 2004). Diapause has been most extensively studied in insects with multivoltine life cycles (e.g., Bao & Xu, 2011; Emerson, Bradshaw, & Holzapfel, 2010; Lehmann et al., 2014; Ragland, Denlinger, & Hahn, 2010; Schmidt et al., 2008). For multivoltine organisms, trophic resources tend to have broad temporal distributions, and selection favours fast resumption of nondiapause development as soon as abiotic conditions are physiologically permissive. Thus, it is not surprising that for multivoltine species, the diapause maintenance phase typically terminates during winter, with insects remaining in post‐diapause quiescence (dormant due to physiologically restrictive abiotic conditions rather than programmed metabolic and developmental suppression) until physiological temperature thresholds are surpassed, after which time they initiate post‐diapause development (Hayward, Pavlides, Tammariello, Rinehart, & Denlinger, 2005). In contrast, univoltine insects with narrow resource windows like R. pomonella derive no benefit from terminating diapause maintenance early during the winter because they must remain dormant even after conditions become permissive the following spring until it is time to initiate the post‐diapause developmental phase to synchronize themselves with host fruit phenology.

The response of R. pomonella to the end of simulated winter was therefore more metabolically dynamic than we expected (Figure 5). The shift from overwinter respiration to post‐winter diapause respiration was neither a gradual linear increase nor a discrete stair‐step up to the post‐winter level. Rather, pupae showed a rapid but transient metabolic increase immediately after removal from the cold, peaking around 48 hr at a level 4‐ to 5‐fold higher than the eventual post‐winter baseline metabolic rates (Figure 5). During normal development, metabolic rates of ectotherms increase with temperature (Angilletta, 2009). Thus, thermally sensitive metabolic rates can lead to increased energy consumption during warmer conditions, which may prove costly to univoltine insects that need to conserve energy stores for prolonged dormancy (Irwin & Lee, 2000; Koštál et al., 2011; Mercader & Scriber, 2008; Vrba et al., 2014). However, diapausing insects may be able to mitigate these energetic costs by reducing the thermal sensitivity of their metabolic rates (Williams, Chick, & Sinclair, 2015; Williams et al., 2012), as we saw in our trials when initially high metabolic rates were reduced within days of transfer to warm, permissive thermal conditions. Active suppression of metabolism in the face of physiologically permissive temperatures may play an important role in maintaining optimal overwintering energetics in the face of intermittent periods of warmth both in the fall and spring (Hahn & Denlinger, 2007; 2011; Sinclair, 2015). Differences in the immediate post‐winter metabolic dynamics of apple and hawthorn flies might therefore represent an additional axis of divergent seasonal adaptation in this classic speciation system, warranting further study.

CONCLUSIONS

Specialist insects make up a large proportion of metazoan biodiversity (Bush & Butlin, 2004; Jaenike, 1990). How insect life‐history timing evolves has important implications for both the origin and maintenance of a large proportion of terrestrial biodiversity (Bush, 1993). The tempo of seasonal adaptation for many species likely depends on the covariance structure of different components of dormancy regulation and how these components act to influence phenology. Are different facets of diapause development free to evolve independently, or are shifts in life‐history timing constrained by a need for simultaneous evolution across several phases of diapause development? Here, we found that termination of the diapause maintenance phase plays a particularly important (but not sole) role in synchronizing insect phenology with that of ephemeral resources (Figures 3 and 4). In the case of R. pomonella, complex patterns of quantitative allelic variation appear to act primarily through the hub of diapause termination to affect seasonal timing, with additional input via post‐diapause development. The dual roles of these two developmental phases may be a coarse and fine adjustment of phenology, respectively. It is important to note that the magnitude of effects of diapause termination and post‐diapause development reported here were measured under a simplified laboratory temperature regime: a constant post‐winter temperature of 25°C. It remains to be determined whether the effects of either of these phases on eclosion phenology are themselves function‐valued traits driven by temperature. One possibility is that the duration of the diapause maintenance phase sets life‐history timing broadly across years and that the post‐diapause development module may allow fine‐tuning of adult emergence timing due to interannual variation in seasonal conditions in host plant phenology. For example, apples always fruit earlier than hawthorns in our sympatric field sites. However, in some warm years, apples and hawthorns both fruit earlier and in cooler years both fruit later. Metabolic rates during the post‐diapause developmental phase are substantially higher than during the diapause maintenance phase and we speculate that post‐diapause development may be more responsive to temperature acting to fine‐tune synchrony with host plant availability and local abiotic conditions from year to year at any particular location and across a latitudinal gradient of seasonal conditions throughout this fly's range. Future work on the thermal sensitivity of specific modular components across these two modules of the diapause developmental trajectory is needed to test this hypothesis for synchronizing adult emergence timing with weather‐related local variation in host fruit phenology.

In Rhagoletis flies, variation in seasonal adult emergence timing is associated with ongoing divergent adaptation to the novel apple host plant and reproductive isolation, but this shift to earlier seasonal timing is analogous to expectations for phenological shifts under climate change in coming decades. A primary motivation of this research was to determine the extent to which the inherent complexity of diapause development is likely to aid or constrain the rapid evolution of seasonality in univoltine insects. In the case of these two host races of R. pomonella, the partially independent major and minor effects related to the timing of the transition out of diapause maintenance and components of post‐diapause development bodes well for the continued evolvability of adult eclosion time in this system. If phenological adaptation in these flies is not primarily constrained by strong antagonistic pleiotropy among different facets of diapause development, it may be limited more by available genetic variation affecting each intermediate phenotype. However, the apple and hawthorn fly speciation story is superimposed on strong local phenological adaptation of these flies along latitudinal clines as well (Dambroski & Feder, 2007; Doellman, Egan, et al., 2018; Doellman, Ragland, et al., 2018; Michel et al., 2010), which may act as a vast reservoir for standing genetic variation in seasonality traits across the continent (Powell et al., 2013). The combination of relatively simple, major effect basis for variation in adult emergence phenology coupled with abundant genetic variation may mean that natural selection has a fairly free hand with these insects to keep pace with changing seasonal conditions or colonize novel niches. However, the existence of moderate antagonistic covariance among traits with links to phenology may potentially constrain the evolution of novel phenotypes outside the current trait distribution. Furthermore, elucidating the roles of diapause maintenance and post‐diapause development as important intermediate phenotypes driving differences in phenology sets the stage for future studies to begin drawing stronger connections between allelic variation across the genome and the complex ecological traits of life‐history timing, local adaptation and ecological reproductive isolation. Given that genetic variation in the diapause maintenance phase (Wadsworth et al., 2013) or in post‐diapause development (Posledovich et al., 2015; Posledovich et al., 2014; Stålhandske, Gotthard, Posledovich, & Leimar, 2014) is thought to be independent drivers of adaptive shifts in phenology in other insect systems, we expect that the dual role for diapause and post‐diapause development in regulating life‐history timing may be partially modular across a wide range of insect species. Thus, the idea that phenologies may be tuned to local seasonality by a combination of genetic and environmental effects (G×E) is broadly applicable across diverse groups of insects and germane to understanding persistence versus declines of insect populations in the future as seasonality is further altered by anthropogenic change.

ACKNOWLEDGMENTS

We would like to thank Meredith Doellman, Glen Hood and Pete Meyers for help with field collections. Caroline Williams provided invaluable advice on our respirometry configuration. Help with phenotyping was provided by Jennifer Serviss, Caelen Schreiber, Andre Szejner, Chao Chen, Genevieve Comeau and Xiaoping Wang. This research was supported by NSF IOS 170773 to GJR and JLF, NSF DEB 1638997 to GJR, NSF DEB 1638951 to JLF and NSF DEB 1639005 to THQP and DAH, as well as NSF IOS 1257298, the Florida Agricultural Experiment Station, and the joint FAO/IAEA CRP Dormancy Management to Enable Mass‐rearing to DAH.

DATA AVAILABILITY STATEMENT

All data are available in DRYAD: Powell et al. (2020), Respirometry data for “A rapidly evolved shift in life‐history timing during ecological speciation is driven by the transition between developmental phases,” Dryad, Dataset, https://doi.org/10.5061/dryad.8cz8w9gmq.

REFERENCES